Abstract
We recently discovered an adaptable contrast comparison process (which we sometimes call “Buffy adaptation”). At each spatial position in a visual pattern, a contrast comparison level continually and rapidly (less than 1 sec) updates to equal the recent average contrast. The current contrast is then processed relative to that level, without regard to sign (in a second-order task such as ours). This process acts in concert with a previously known contrast-gain control of the normalization type.
The patterns in the current experiments are 2x2 grids of Gabor patches (windowed sinusoidal grating patches; approximately 2 cpd; presented at fixation). The adapt pattern - a grid of Gabors all at the same contrast - is shown for 1 sec, and then a test pattern is shown for 94 msec, and then the adapt pattern is shown again for 1 sec. The test pattern is horizontal (or vertical) contrast-defined stripes: the Gabors in the first row (or column) have contrast C1 and those in the second row (or column) have contrast C2. The task of the observer is to identify the orientation of these stripes. The results curves have a “butterfly” shape (like an upside-down ‘W’) when plotted as percent correct identification versus average test contrast (average of C1 and C2). When the average test contrast equals the adapt pattern's contrast, performance is poor (butterfly's center). If the average test contrast is increased or decreased while keeping the difference between the two test contrasts constant (e.g. |C2−C1|=10%), performance improves (butterfly's wing's inner edges) and then declines again (butterfly's wing's outer edges).
These results are an interaction of (1) the adaptable contrast comparison process, producing the poor performance at the butterfly's center, and (2) the normalization type process, producing the declines on the outside edges of the butterfly's wings.
Supported in part by NIH grant EY08459.